Module python_sandbox 

Modern Python Plugin Runtime

AI-era design for running Python plugins in SochDB using:

• Pyodide: Full CPython 3.12 with numpy, pandas, scikit-learn
• WASM Component Model: Standard interfaces for cross-language composition
• AI Triggers: Natural language → Python code generation

Architecture

┌──────────────────────────────────────────────────────────┐
│                  Python Plugin System                     │
├──────────────────────────────────────────────────────────┤
│  PythonPlugin → PyodideRuntime → WASM Sandbox            │
│       ↓              ↓                ↓                  │
│   packages:      micropip          Memory isolation      │
│   numpy,pandas   install           Resource metering     │
└──────────────────────────────────────────────────────────┘

Example

let runtime = PyodideRuntime::new(RuntimeConfig::default()).await?;
runtime.install_packages(&["numpy", "pandas"]).await?;

let plugin = PythonPlugin::new("fraud_detector")
    .with_code(r#"
        import numpy as np
        def on_insert(row):
            if row["amount"] > 10000:
                raise TriggerAbort("Amount too high")
            return row
    "#)
    .with_trigger("transactions", TriggerEvent::BeforeInsert);

runtime.register(plugin)?;

Structs

AiTriggerGenerator

Generates Python trigger code from natural language instructions

PyodideRuntime

Pyodide-based Python runtime

PythonPlugin

A Python plugin with code and trigger bindings

RuntimeConfig

Configuration for the Pyodide runtime

RuntimeStats

Runtime statistics

TriggerContext

Context passed to trigger execution

Enums

TriggerEvent

Types of trigger events

TriggerResult

Result from trigger execution